Eye Tracking: Challenging the Biological Monopoly

Here at Eye Tracking Update we’re often talking about eye tracking in humans, and thus dealing with humans’ physiological processes of seeing. But as we know, engineers are attempting to design other, less human objects with an ability to “see”, threatening a living organism’s monopoly on sight. Computers now have image tracking features and object recognition technology (see Google’s photo recognition software for an example). What exactly, you might ask, am I talking about here?

Check out a new experiment that was recently brought to our attention involving neural processing of natural environments. Scientists at Harvard and MIT have led new studies attempting to reverse engineer the biological visual system, in an effort to further understand the human brain and build an artificial system that works in a similar way to a living organism’s vision. In a recent study led by David Cox, Principal Investigator of the Visual Neuroscience Group at Harvard University and Nicolas Pinto, Ph.D. Candidate in Brain and Cognitive Sciences at MIT, the two attempted to do just that, publishing their findings last year in a journal of biology.

The team spent time creating “biologically inspired algorithms that enable computers to understand what they see.” Their goal was to approximate the ease with which humans see the world and apply that to computers’ vision, but as of yet, there are no computers that come close to the sophistication with which the human brain’s visual process works. Image tracking is becoming more common in computers, but the extent of this technology based on a computer’s interpretation of a scene before it, made up of just a collection of numbers. The scientists’ ultimate goal is to create computers that “see” in the same way humans do, extracting structure and meaning from images in the same way we do.

Biological seeing works when light from a scene enters the retina and is converted to information with the help of light-detecting cells in the back of the eye. That information is passed along the optic nerve to the cerebral cortex. From there, over a hundred million neurons, each of which transfers the input information to output, each working like a tiny and powerful computer. From that point, there are various stages of processing before the information is interpreted and meaning applied.

The scientists at Harvard and MIT are attempting to simulate the hierarchical fashion that information processing is organized in the visual cortex. They plan to create an artificial system utilizing simulated neurons that process input information based on mathematical algorithms. For this, quite a number of supercomputers are needed in order to approximate the sophistication and computing power of the human brain. But simply lining up large supercomputers to match the power of the brain’s processing input/output isn’t enough, these computers need to be organized in a fashion so that they all work together to do what our brains can do.

The key to the scientists’ experiment, however, was to build thousands of candidate models, instead of just one prototype. Each one was screened for those that performed best on an object recognition task. With the resulting models outperforming state-of-the-art computer vision recognition systems, the scientists’ designs were able to accurately identify a range of objects in their natural environments. But they wouldn’t have been able to do it without modern graphics hardware, which enabled them to carry out the study for a fraction of the cost it would have taken using ordinary computer processing units.

They hope to apply their findings to other areas of simulated vision, face recognition, object and eye tracking, and gesture and action recognition among others. Check the link out below for a video presentation by David Cox.

Researchers demonstrate a better way for computers to ‘see’ (w/ Video)

Related articles:

1. Eye Tracking Research Measures Effectiveness of Super Bowl Ads
2. Eye Tracking on the Cheap: Making MacGyver Proud
3. Tips From Eye Tracking Studies on Website Design
4. Eye Tracking: Nose Tracking (Update?)
5. Eye Tracking: Investigating What Women Notice in Advertisements
6. Using Eye Tracking to Study the “Humor Effect”
7. Eye Tracking: Facebook and LinkedIn Usability
8. Introduction to Eye Tracking: The Anatomy of the Eye
9. Eye Controlled Video Games? Better Late Than Never
10. Pupil Tracking in Airport Security: Can Body Language Indicate Terrorist Intent?